The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted content from external GitHub issues.
Ingestion points: The skill uses gh issue view <ref> --json title,body,comments in SKILL.md to fetch external data.
Boundary markers: While it instructs the agent to "quote" the issue, it lacks explicit delimiters or warnings to ignore instructions embedded within the issue body.
Capability inventory: The skill possesses significant capabilities, including shell command execution (curl, gh, jq), creating/deleting sandboxes via the Veris API, executing local repository scripts (.veris/run.sh), and opening pull requests.
Sanitization: There is no evidence of sanitization or filtering of the ingested issue data.
[COMMAND_EXECUTION]: The skill performs various shell operations to interact with local and remote resources.
Evidence: It utilizes curl for API communication with svc.api.veris.ai, jq for data manipulation, and executes a local script .veris/run.sh to trigger the application's code path. These actions are necessary for the skill's stated purpose but rely on the integrity of the repository and the vendor's infrastructure.

fix